Can We Build a Monolithic Model for Fake Image Detection? SICA: Semantic-Induced Constrained Adaptation for Unified-Yet-Discriminative Artifact Feature Space Reconstruction

Bo Du, Xiaochen Ma, Xuekang Zhu, Zhe Yang, Chaoqun Niu, Jian Liu, Ji-Zhe Zhou
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:26529-26551, 2026.

Abstract

Fake Image Detection (FID), aiming at unified detection across four image forensic subdomains, is critical in real-world forensic scenarios. Compared with ensemble approaches, monolithic FID models are theoretically more promising, but to date, consistently yield inferior performance in practice. In this work, we identify the intrinsic distinctness of artifacts across subdomains—a critical barrier we term the "Ji-Zhe phenomenon". Driven by this phenomenon, we diagnose the cause of this underperformance for the first time: the collapse of the artifact feature space. The core challenge for developing a practical monolithic FID model thus boils down to the "unified-yet-discriminative" reconstruction of the artifact feature space. To address this paradoxical challenge, we hypothesize that high-level semantics can serve as a structural prior for the reconstruction, and further propose Semantic-Induced Constrained Adaptation (SICA), the first monolithic FID paradigm. Extensive experiments on our $ \textit{OpenMMSec} $ dataset demonstrate that SICA outperforms 15 state-of-the-art methods and reconstructs the target unified-yet-discriminative artifact feature space in a near-orthogonal manner, thus firmly validating our hypothesis. The code and dataset will be made publicly available.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-du26g, title = {Can We Build a Monolithic Model for Fake Image Detection? {SICA}: Semantic-Induced Constrained Adaptation for Unified-Yet-Discriminative Artifact Feature Space Reconstruction}, author = {Du, Bo and Ma, Xiaochen and Zhu, Xuekang and Yang, Zhe and Niu, Chaoqun and Liu, Jian and Zhou, Ji-Zhe}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {26529--26551}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/du26g/du26g.pdf}, url = {https://proceedings.mlr.press/v306/du26g.html}, abstract = {Fake Image Detection (FID), aiming at unified detection across four image forensic subdomains, is critical in real-world forensic scenarios. Compared with ensemble approaches, monolithic FID models are theoretically more promising, but to date, consistently yield inferior performance in practice. In this work, we identify the intrinsic distinctness of artifacts across subdomains—a critical barrier we term the "Ji-Zhe phenomenon". Driven by this phenomenon, we diagnose the cause of this underperformance for the first time: the collapse of the artifact feature space. The core challenge for developing a practical monolithic FID model thus boils down to the "unified-yet-discriminative" reconstruction of the artifact feature space. To address this paradoxical challenge, we hypothesize that high-level semantics can serve as a structural prior for the reconstruction, and further propose Semantic-Induced Constrained Adaptation (SICA), the first monolithic FID paradigm. Extensive experiments on our $ \textit{OpenMMSec} $ dataset demonstrate that SICA outperforms 15 state-of-the-art methods and reconstructs the target unified-yet-discriminative artifact feature space in a near-orthogonal manner, thus firmly validating our hypothesis. The code and dataset will be made publicly available.} }
Endnote
%0 Conference Paper %T Can We Build a Monolithic Model for Fake Image Detection? SICA: Semantic-Induced Constrained Adaptation for Unified-Yet-Discriminative Artifact Feature Space Reconstruction %A Bo Du %A Xiaochen Ma %A Xuekang Zhu %A Zhe Yang %A Chaoqun Niu %A Jian Liu %A Ji-Zhe Zhou %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-du26g %I PMLR %P 26529--26551 %U https://proceedings.mlr.press/v306/du26g.html %V 306 %X Fake Image Detection (FID), aiming at unified detection across four image forensic subdomains, is critical in real-world forensic scenarios. Compared with ensemble approaches, monolithic FID models are theoretically more promising, but to date, consistently yield inferior performance in practice. In this work, we identify the intrinsic distinctness of artifacts across subdomains—a critical barrier we term the "Ji-Zhe phenomenon". Driven by this phenomenon, we diagnose the cause of this underperformance for the first time: the collapse of the artifact feature space. The core challenge for developing a practical monolithic FID model thus boils down to the "unified-yet-discriminative" reconstruction of the artifact feature space. To address this paradoxical challenge, we hypothesize that high-level semantics can serve as a structural prior for the reconstruction, and further propose Semantic-Induced Constrained Adaptation (SICA), the first monolithic FID paradigm. Extensive experiments on our $ \textit{OpenMMSec} $ dataset demonstrate that SICA outperforms 15 state-of-the-art methods and reconstructs the target unified-yet-discriminative artifact feature space in a near-orthogonal manner, thus firmly validating our hypothesis. The code and dataset will be made publicly available.
APA
Du, B., Ma, X., Zhu, X., Yang, Z., Niu, C., Liu, J. & Zhou, J.. (2026). Can We Build a Monolithic Model for Fake Image Detection? SICA: Semantic-Induced Constrained Adaptation for Unified-Yet-Discriminative Artifact Feature Space Reconstruction. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:26529-26551 Available from https://proceedings.mlr.press/v306/du26g.html.

Related Material